symcli control operations with srdf family products

Engineering White Paper

Using SYMCLI to Perform Control Operations with SRDF Family Products

Abstract

This paper provides an introduction to the SRDF functionality that allows you to transmit copies of the data from a Symmetrix unit located at the production site to a remotely located Symmetrix unit.

Published 4/5/2004

4/5/2004

Copyright 2004 EMC Corporation. All rights reserved.

EMC believes the information in this publication is accurate as of its publication date. The information is subject to change without notice.

THE INFORMATION IN THIS PUBLICATION IS PROVIDED AS IS. EMC CORPORATION MAKES NO REPRESENTATIONS OR WARRANTIES OF ANY KIND WITH RESPECT TO THE INFORMATION IN THIS PUBLICATION, AND SPECIFICALLY DISCLAIMS IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Use, copying, and distribution of any EMC software described in this publication requires an applicable software license.

Part Number 300-000-076 REV H H604.6

Using SYMCLI to Perform Control Operations with SRDF Family Products 2

4/5/2004

Table of Contents

Introduction.........................................................................................................5 Purpose and Scope ..................................................................................................................... 5 Related Documentation ............................................................................................................... 5

Practical Uses .....................................................................................................6

SRDF Configurations and Methods of Implementation...................................8 RDF Groups................................................................................................................................. 9

Setting Up Device Groups and Composite Groups .........................................9

Invalid Tracks....................................................................................................10

SRDF Control Operations ................................................................................11

SRDF Pair States...............................................................................................12

Suspending the RDF Links of an SRDF Pair ..................................................13

Establishing an SRDF Pair...............................................................................14

Splitting an SRDF Pair......................................................................................15

Restoring an SRDF Pair ...................................................................................16

Failover and Failback .......................................................................................16 Failback...................................................................................................................................... 17 Failover with a Fast Swap and Establish ................................................................................... 17

Updating the R1 Device....................................................................................18

Using Concurrent RDF Devices.......................................................................19

Swapping R1 Devices with R2 Devices...........................................................21

Data Mobility Replication .................................................................................22

SRDF Asynchronous (SRDF/A) Replication ...................................................23


4/5/2004

Using BCVs to Preserve a Copy of the Remote SRDF/A Data ................................................. 24 Setting Up SRDF Asynchronous Operation............................................................................... 25

Performing SRDF Control Operations in Parallel...........................................26

Creating Dynamic SRDF Pairs .........................................................................27

Creating Dynamic RDF Groups .......................................................................29

Using SRDF and TimeFinder/Mirror in Multi-Hop Configurations ................30 Establishing a BCV Pair with an RDF BCV ............................................................................... 30 Copying Data to a Remotely-Associated BCV........................................................................... 31 Copying Data in a Complex Multi-Hop Environment ................................................................. 32

Targeting Remote Devices When the Data Source Is the Local Standard RDF Device ....... 33 Targeting Remote Devices When Remote Copying Is from a Local R1 BCV ....................... 34

Using a Composite Group to Control a Set of Devices That Spans Multiple Symmetrix Units................................................................................................36

Example 1: Basic SRDF Control Operations .................................................38

Example 2: Concurrent RDF ...........................................................................59

Example 3: Creating Dynamic SRDF Pairs .....................................................72

Example 4: Creating a Dynamic RDF Group...................................................78

Example 5: Operating with SRDF Asynchronous Replication......................82

Example 6: Using a Composite Group to Contol SRDF Pairs.......................90

Appendix A: Dynamic RDF with Enginuity Versions 5x67 and Higher ......101


4/5/2004

Introduction The SRDF product family (referred to as SRDF in this document) provides a mirrored data storage solution that allows you to duplicate production site data on one or more remote target Symmetrix systems. When your main systems are down for a planned or unplanned event, SRDF enables fast switchover from the source data to the target copy.

SRDF over IP (Internet Protocol) can create copies of the data and transmit these copies over high-bandwidth IP-based Virtual Private Networks to a remote Symmetrix system. When distance impacts performance, SRDF can use delayed synchronization methods of data replication to ensure data mobility.

SRDF products currently support the following methods of data replication:

Synchronous Replication provides real-time mirroring of data between the source Symmetrix and the target Symmetrix systems. Data is written simultaneously to the cache of both systems in real time before the application I/O is completed, thus ensuring the highest possible data availability.

Semi-Synchronous Replication writes data to the source system, completes the I/O, and then synchronizes the data with the target system. Since the I/O is completed prior to synchronizing data with the target system, this method provides an added performance advantage. A second write will not be accepted on a Symmetrix source device until its target device has been synchronized.

Adaptive Copy Replication transfers data from the source devices to the remote devices without waiting for an acknowledgment. This is especially useful when transferring large amounts of data during data center migrations, consolidations, and in data mobility environments.

Asynchronous Replication places host writes into cycles or chunks and then transfers the entire chunks to the target system. When a complete chunk is received on the R2 side, the copy cycle is committed. If the RDF links are lost during data transfer, any partial chunk is discarded, preserving consistency on the R2. Beginning with Solutions Enabler version 5.3 running on Symmetrix units using Enginuity version 5670, this method provides a consistent point-in-time R2 image that is not far behind the R1 side and results in minimal data loss if there is a disaster at the source site.

All methods of replication can co-exist in a Symmetrix unit, allowing you to specify the method on a per-device basis. No special application coding is required and no CPU overhead is incurred.

The local SRDF device, known as the source (R1) device, is configured in a pairing relationship with a remote target (R2) device, forming an SRDF pair. While the R2 device is mirrored with the R1 device, the R2 device is write disabled or not ready to its host. (Not ready means disabled for both reads and writes.) After the R2 device becomes synchronized with its R1 device, you can split the R2 device from the R1 device at any time, making the R2 device fully accessible again to its host. After the split, the target (R2) device contains valid data and is available for performing business continuance tasks through its original device address or restoring (copying) data to the source (R1) device if there is a loss of data on that device.

Purpose and Scope

This document describes SRDF functionality in versions of EMC Solutions Enabler up to version 5.4 running on Symmetrix units using Enginuity versions 5x65 to 5x67, 5568, 5669, and 5670.

Related Documentation

The following documents provide information related to the concepts discussed in this paper:

EMC Solutions Enabler Symmetrix SRDF CLI Product Guide

EMC Solutions Enabler Symmetrix Base Control Product Guide

EMC Solutions Enabler Symmetrix TimeFinder CLI Product Guide


4/5/2004

Using SYMCLI to Query and Verify with SRDF Family Products (P/N 300-000-077)

Using SYMCLI to Implement RDF Consistency Protection with SRDF Family Products (P/N 300-000-284)

Using SYMCLI to Set Up TimeFinder/Mirror BCV Pairs (P/N 300-000-072)

Using SYMCLI to Perform TimeFinder/Mirror Control Operations (P/N 300-000-074)

Using SYMCLI to Perform SRDF/AR (P/N 300-000-078)

Practical Uses Practical uses of suspend and resume operations usually involve unplanned situations in which you want to immediately suspend I/O between the R1 and R2 devices over the RDF links. In this way, you can stop any data propagation problems or perform immediate backups without affecting I/O from the local host application. You can then resume I/O between the R1 and R2 and return to normal operation.

Practical uses of establish and split operations usually involve planned situations in which you want to use the R2 copy of the data without interfering with normal write operations to the R1 device. Splitting a point-in-time copy of data allows you to access that data on the R2 device for various business continuance tasks. The ease of splitting SRDF pairs into separate database instances providing exact copies makes it convenient to perform scheduled backup operations, data warehousing, or new application testing from the target Symmetrix data while normal operations continue on the source Symmetrix.

You can also use the R2 copy to test disaster recovery plans without manually intensive recovery drills, complex procedures, and business interruptions. You can also test upgrades to new versions or change actual code without affecting your online production server. Simply run the experimental server on the R2 copy of the database until the upgraded code runs with no errors. Then upgrade the production server.

In cases where an absolute realtime database is not essential, you can split the SRDF pair periodically and use the R2 copy for queries and report generation. Then you can re-establish the SRDF pair periodically to provide incremental updating of data on the R2 device. The ability to refresh the R2 device periodically allows you to provide the latest information for data processing and reporting.

Practical uses of failover and failback operations usually involve situations in which you need to switch business operations from the production site to a remote site. Once the switch occurs, normal operations continue using the remote (R2) copy of synchronized application data. Scheduled maintenance at the production site is one reason you might want to failover to the R2 site. Scheduled maintenance can include such tasks as operating system upgrades, host processor upgrades, and environmental disruptions.

Disaster recovery is another reason to temporarily failover to a remote site. The typical recovery routine involves customized software and complex procedures. Offsite media must be either electronically transmitted or physically shipped to the recovery site. Time-consuming restores and the application of logs usually follow. SRDFs failover/failback operations significantly reduce the restore time by incrementally updating only the specific tracks that have changed, accomplishing in minutes what might take hours for a complete load from a dumped database volume. Moreover, you can start the server and run it to its full production capability while the synchronization is still in progress.

Practical uses of the R1 update operation usually involve situations in which you want the R1 to become almost synchronized with the R2 before a failback, while the R2 side is still online to its host. You can include the until track_threshold option to identify a number of invalid tracks that can build up from the active local I/O on the R2 side before retriggering another update operation. Note, however, that SYMCLI does not retrigger another update until the previous batch of updates is fully copied to the R1 side.


4/5/2004

Concurrent RDF means having two target R2 devices configured as concurrent mirrors of one source R1 device. Using a concurrent SRDF pair like this allows you to easily generate two copies of the same data at remote locations. When you split the two R2 devices from their source R1 device, each target sites host can access its own data.

Swapping R1/R2 devices of an SRDF pair causes the source R1 device to become a target R2 device, and the former target device becomes the source device. Swapping SRDF devices allows the R2 side to take over operations while retaining a remote mirror on the R1 side. Swapping is especially useful after failing over an application from the R1 side to the R2 side. Dynamic Swap (a very fast version) is available with Enginuity version 5567 or higher.

Data Mobility replication is an SRDF configuration that restricts SRDF devices to operating only with Adaptive Copy replication. This is useful when you need to ensure that certain devices are used only for transferring large amounts of data in data mobility environments.

Parallel processing allows you to perform SRDF operations in a less time-consuming manner than performing the same operations sequentially, especially establish operations that need to mark and merge track tables between the R1 and R2 devices. If an application requires establishing multiple device groups across multiple Symmetrix units, doing so in parallel can significantly reduce the time required for the operation to complete. Beginning with the SRDF component of EMC Solutions Enabler version 5.0, if you want to perform SRDF parallel processing within a single Symmetrix unit, you can perform up to sixteen parallel control operations within a single Symmetrix unit across sixteen different RDF groups. (With EMC Solutions Enabler version 4.3, you can perform up to four parallel control operations within a single Symmetrix unit.) Beginning with Solutions Enabler version 5.2 running on Symmetrix units using Enginuity version 5669, parallel processing is controlled at the device level rather than at the RDF group level, allowing you to perform up to 64 parallel processing operations on different devices.

Dynamic SRDF allows you to create your own dynamic SRDF pairs from non-SRDF devices while the Symmetrix unit is in operation (on the fly). Historically, source and target SRDF device pairing has been static once set at configuration time, and this is still true of devices that are configured as non-dynamic SRDF devices. However, beginning with the SRDF component of EMC Solutions Enabler version 5.0 running on Symmetrix units using Enginuity version 5568, you now have the capability of using RDF-capable, non-SRDF devices in creating and synchronizing SRDF pairs. This feature provides greater flexibility in deciding where to copy protected data.

Beginning with Solutions Enabler version 5.2 running on Symmetrix units using Enginuity version 5669, you can create dynamic RDF groups in a Switched Fabric SRDF environment. An RDF group (RA group) represents a logical connection between two Symmetrix units. Historically, RDF groups were limited to those static RDF groups defined at configuration time. However, you can now create, modify, and delete RDF groups while the Symmetrix unit is in operation. This feature provides greater flexibility in forming SRDF-pair-associated links.

SRDF Asynchronous replication (SRDF/A) allows you to maintain a consistent point-in-time R2 image with only a slight lag behind the R1 side, which results in minimal data loss if there is a break in communication between the source and target. Beginning with Solutions Enabler version 5.3 running on Symmetrix units using Enginuity version 5670, SRDF/A allows you to configure an SRDF environment in which R1 devices transfer data to R2 devices in chunk cycles. One difference between Asynchronous replication and other methods of replication is that less data is transferred, and the data is handled fewer times. If the same data is updated multiple times in the same cycle (for example, the same track is written to ten times in a given time period), that data is sent across the RDF links only once and does not have to be copied within the cache for each update as in other ordered write solutions. This performance benefit can result in more efficient link utilization and potentially lower link bandwidth requirements when compared to a traditional ordered write solution that must handle each update write discretely. This saving in cycle size is most relevant for applications that can tolerate a loss of up to twice the size of a cycle if the RDF links are lost.


4/5/2004

SRDF Configurations and Methods of Implementation An SRDF configuration has at least one source unit and one target unit. SRDF configurations may transfer data in a uni-directional or bi-directional manner. In a uni-directional configuration, all R1 devices reside in the source Symmetrix unit and all R2 devices in the target Symmetrix unit. Under normal conditions, data flows from the R1 devices to the target R2 devices. In a bi-directional configuration shown in Figure 1, both R1 and target R2 devices reside in each Symmetrix unit. Data flows from the source (R1) devices in their respective Symmetrix unit to their corresponding target (R2) devices in the other Symmetrix unit.

SymmetrixSite A

SymmetrixSite B

CLI-000104

Host Host

RDF Links

SRDF PairSource

(R1)Device

Target(R2)

DeviceI/O Transfer

SRDF PairTarget(R2)

Device

Source(R1)

DeviceI/O Transfer

Figure 1. SRDF Bi-Directional Configuration

SRDF connectivity implementations encompass several solutions, depending on geographical requirements. The ESCON-based Campus Solution provides mirroring operations for Symmetrix units located up to 60 kilometers (37.5 miles) apart using fiber-optic links. SRDF over Fiber Channel topology supports high-speed synchronous mirroring among Symmetrix systems in campus situations.

With the Extended Distance Solution for sites farther removed geographically, SRDF FarPoint software supports mirroring across SRDF-supported telecommunications links, including T1/E1, T3/E3, and ATM. For businesses that have an intranet based on IP, SRDF over IP is another solution.

An SRDF topology can incorporate open network switching (fabric) in the RDF links. The switched RDF involves non-blocking switching devices that interconnect two or more nodes. Symmetrix units in a switched RDF topology can have each port pair running full-duplex.

During configuration, SRDF pairs are usually configured such that the R1 device and its paired R2 device are the same size. Although it is possible to reconfigure devices and migrate data from an R1 device to a larger size R2 device, your ability to access that data may require file system operations on the target-side host. For information about reconfiguring existing devices, refer to the white paper Using the SYMCLI Configuration Manager (P/N 300-000-475).


4/5/2004

RDF Groups

The Remote Link Directors (RLDs), which manage the data transfers between Symmetrix units, have either an RA1 or RA2 designation. RA1s reside in the source Symmetrix, RA2s in the target Symmetrix. These RLDs have to be assigned to an RDF group as part of each Symmetrix configuration. An RDF group (also called an RA group) is a logical grouping of RDF devices that are serviced by the same set of RLDs.

RDF groups are related to physical RDF connections. Each link is logically associated with an RDF group at the time the Symmetrix is configured. An RDF group is configured and assigned an RDF group number by Enginuity. Typically, there are two or more RA/RF directors per RDF group.

The symcfg list command displays all Symmetrix units attached to your host and reachable via RDF links. Adding the RA all option to the command displays how many RA/RF directors exist.

symcfg list symcfg list RA all

Setting Up Device Groups and Composite Groups During configuration, SRDF devices are configured in pairing relationships and usually established so that the SRDF pairs mirroring activity becomes operational as soon as the RDF links are online. The Symmetrix global memory stores information about the pair state of operational SRDF devices.

Unlike the RDF group described in the previous section, the device group and the composite group are entities that you can create and use to manage and control SRDF pairs. Your hosts SYMAPI1 database file or the GNS-managed global repository stores information about a device group or composite group and the devices contained in the group.

Beginning with Solutions Enabler version 5.4, you can create a composite group to control a set of SRDF pairs and BCV pairs2 that spans multiple Symmetrix units. When an RDF composite group is enabled for consistency protection, it is known as an RDF consistency group. A composite group provides greater flexibility than a device group, which can define devices only on a single Symmetrix unit. However, unlike the device group, the composite group cannot currently operate on specific pairs within the group but must perform an operation on the entire group. For more information about composite groups, refer to the section Using a Composite Group to Control a Set of Devices That Spans Multiple Symmetrix Units.

An SRDF device group or composite group can hold one of two types of standard devices: RDF1 (source) or RDF2 (target). An SRDF device (R1 or R2) can be moved to another device group or composite group only if the source and destination groups are the same group type.

You can build an RDF1 type group on a host attached to a Symmetrix that contains those RDF1 devices. If a host is attached to a Symmetrix that contains RDF2 devices, you can build an RDF2 type group on that host. You can perform the same SRDF operations from any attached host that contains the group definition. For information about propagating group definitions globally, refer to the white paper Using SYMCLI and GNS to Propagate Group Definitions to Multiple Hosts (P/N 300-000-384).

When adding RDF standard devices to a device group or composite group, all devices in the device group:

Must be SRDF devices

Must be either all RDF1 or RDF2 type devices, as specified by the group type

1 Symmetrix API

2 In versions of Solutions Enabler prior to version 5.4, you can use a consistency group to control SRDF pairs only.


4/5/2004

Must belong to the same Symmetrix RDF group number

When adding TimeFinder/Mirror RDF BCV devices to a device group or composite group:

You can associate RDF BCV devices with any group type, including the Regular type

You can choose RDF BCV devices that belong to a different RDF group number than the SRDF standard devices

You cannot mix RDF1 and RDF2 type RDF BCV devices in the same group

You can associate RDF BCV devices locally, remotely via the SRDF standard devices, or remotely via the locally associated RDF BCV devices

To check the configuration of SRDF devices before adding them to a device group or composite group, you can use the symrdf list command to list SRDF devices configured on Symmetrix units attached to your host:

symrdf list

When you add a device to a device group, a logical device name is assigned to it either by your specifying a logical name on the command line or by default.

The following sequence creates an RDF1 type device group and adds an R1 device to the group:

1. Create a device group named Rdf1Grp:

symdg create Rdf1Grp type rdf1

2. Add an R1 device (Symmetrix device name 085) to the device group on Symmetrix number 000000003264. A default logical name of the form DEV001 is assigned to the R1 device:

symld -g Rdf1Grp sid 3264 add dev 085

Invalid Tracks The concept of invalid tracks in SRDF systems indicates what data is not synchronized between the two devices that form an SRDF pair. On both the source and target sides of an SRDF setup, the Symmetrix keeps an account of the tracks that are owed to the other side. The owed tracks are known as remote invalids.

For example, consider the case of an R1 device whose logical connection to its R2 has been suspended. If both devices are made write accessible, hosts on both sides of the RDF link can write to their respective devices, creating R2 invalid tracks on the R1 side and R1 invalid tracks on the R2 side. Each invalid track represents a track of data that has changed since the two sides were split. To re-establish the logical link between the R1 and R2, the invalid tracks have to be resolved.

Resolution of invalid tracks depends on which operation you perform. For instance, you can have remote invalids on both sides prior to an establish or a restore operation. If so, performing an establish operation indicates to SRDF that you want to copy modified R1 tracks to the R2 side. In the process, any tracks that were modified on the R2 side are overwritten with data from corresponding tracks on the R1 side.

Performing a restore operation indicates the opposite that you want to copy modified R2 tracks to the R1 side. In the process, any tracks that were modified on the R1 side are overwritten with data from corresponding tracks on the R2 side.

For more information on conditions governing composite operations, refer to EMC Solutions Enabler Symmetrix SRDF CLI Product Guide.

It is possible, though not common, to end up with local invalid tracks at the end of a series of singular RDF operations. When that happens, exercise care in determining which data to preserve and which to discard.


4/5/2004

SRDF Control Operations The symrdf command performs the high level control operations of the SRDF environment with two types of low level control operations: composite and singular operations. You perform most SRDF operations using composite operations. A composite operation is made up of several singular operations.

Table 1 lists the singular operations that make up each composite operation.

Table 1. Decomposition of Composite Operations into Singular Operations

Composite Operation

symrdf options

Individual operations When used

Full establish full establish - Write disable R2 devices on RA - Suspend RDF link traffic - Mark target device invalid - Merge track tables - Resume RDF link traffic

- Initial synchronization of RDF mirrors - Replacement of failed drive on the R2 side

Incremental establish

establish - Write disable R2 devices on RA - Suspend RDF link traffic - Refresh tracks on target - Merge track tables - Resume RDF link traffic

Resynchronize RDF mirrors after they have been split and target data can be discarded

Split split - Suspend RDF link traffic - Read write enable R2 to its local host

When both sides need to be independently accessible (e.g., for testing)

Full restore full restore - Write disable R1 to host - Write disable R2 devices on RA - Suspend RDF link traffic - Mark all source tracks invalid - Merge track tables - Resume RDF link traffic - Read write enable R1 to host

- Initial (reverse) synchronization of RDF mirrors - Replacement of failed drive on R1 side

Incremental restore

restore - Write disable R1 to host - Write disable R2 devices on RA - Suspend RDF link traffic - Refresh source invalid tracks - Merge track tables - Resume RDF link traffic - Read write enable R1 to host

Resynchronize RDF mirrors after they have been split and the source data can be discarded

Failover failover - Write disable R1 to hosts - Suspend RDF link traffic - Read write enable R2 to hosts

In the event of a failure of the source site

Failback failback - Write disable R2 on RA - Suspend RDF link traffic - Refresh source invalid tracks (requires use of the force option) - Merge track tables - Resume RDF link traffic - Write enable R1 to hosts

To return to the source site from the target site after the cause of failure has been remedied

Update update - Suspend RDF link traffic - Refresh source invalid tracks (requires use of the force option) - Merge track tables - Resume RDF link traffic

To get the R1 side close to synchronized with the R2 side before a failback, while the R2 side is still online to the host


4/5/2004

SRDF Pair States The SRDF pair state encompasses the state of the R1 device, state of the R2 device, and the status of the RDF links between them. Before you can perform an SRDF control operation successfully, the SRDF pair must be in a state that is valid for that operation.

Table 2 lists the possible SRDF pair states that the symrdf query command can display. Not Ready means read/write disabled. Ready means enabled for both reads and writes. WD is write disabled.

Table 2. SRDF Pair States

SRDF Pair State R1 State RDF Links Status

R2 State Invalid Tracks

SyncInProg Ready Ready Not Ready or WD >0

Synchronized Ready Ready Not Ready or WD 0

Consistent Ready Ready Not Ready or WD 0 (None owed to the R2)

Split Ready Not Ready or WD Ready

Failed Over Not Ready or WD Not Ready Ready

R1 Updated Not Ready or WD Ready or WD Ready 0 (R1 tracks on source)

R1 UpdinProg Not Ready or WD Ready or WD Ready >0 (R1 tracks on source)

Suspended Any status Not Ready or WD Not Ready or WD

Partitioned (from R1) Any status Not Ready Not Available

Partitioned (from R2) Not Available Not Ready Any status

Mixed

Invalid Any status Any status Any status

The Partitioned state means that the Symmetrix API (SymmAPI or SYMAPI) is unable to communicate though the corresponding RDF paths to the remote Symmetrix unit. This state does not necessarily mean that two Symmetrix units have lost communication. For example, if one Symmetrix unit has two RA groups and SYMAPI is unable to communicate to a remote Symmetrix unit via one of those RA groups, only RDF devices belonging to that group are marked Partitioned; RDF devices belonging to the other RA group are not.

The Mixed state appears only with the symdg show display to indicate that there are different SRDF pair states in the device group.

The Invalid state is the default when no other SRDF pair states apply. In this case, the combination of R1, R2, and RDF link statuses do not match any other SRDF pair state. This state can also occur when something goes wrong on the device at the DA level (since symrdf query does not display DA status).

When you initiate an SRDF control operation, SYMAPI checks the state of each SRDF pair involved in the operation. If a pair is not in an SRDF pair state that is valid for that operation, the operation will fail unless you have included the force option with the command. Using the force option with an SRDF control operation forces an SRDF pair to a specified state. For example, the following command initiates a failover operation on all SRDF pairs that are currently in the Split state, which is not a typical state for failover:

symrdf g Rdf1Grp force failover

For more information about which SRDF control operations can use the force option, refer to EMC Solutions Enabler Symmetrix SRDF CLI Product Guide.


4/5/2004

Suspending the RDF Links of an SRDF Pair The only singular control operations that you are likely to use on a routine basis are symrdf suspend and symrdf resume. The suspend operation suspends I/O traffic on the RDF links between one or more SRDF pairs in a device group. The resume operation re-establishes the RDF links for those SRDF pairs that were suspended.

Suspending I/O between the R1 and R2 devices is useful if one or more R1 devices cannot propagate data to its corresponding R2 device. By suspending all data propagation from the R1 devices, you ensure that all data flow to the R2 side is instantly and completely halted. By doing this, you ensure that a consistent database (up to the point in time of the data propagation failure) exists on the remote side of the configuration. This enables applications to still use the remote database.

While the RDF links between an SRDF pair are suspended, local I/O to the R1 devices can still occur. While these updates are not sent to the R2 devices immediately, the system does propagate these updates to the R2 side once you initiate a resumption of normal SRDF mirroring activity.

To initiate a suspend operation, an SRDF pair must be in one of the following states (use the symrdf query command to check the state of SRDF pairs in a device group):

Synchronized

R1 Updated

SyncInProg and the symforce option was specified3

R1 UpdInProg and the symforce option was specified3

You can initiate the suspend from either host. The following command initiates a suspend operation on all SRDF pairs in the device group named Rdf1Grp:

symrdf g Rdf1Grp suspend

To resume the RDF links between an SRDF pair, the pair must be in the Suspended state. Assuming that all SRDF pairs in the device group are in the Suspended state, the following command resumes I/O traffic between all SRDF pairs in the device group.

symrdf g Rdf1Grp resume

There are many reasons why a track table may need to be merged while an SRDF pair is in the Suspended state. If the symrdf resume command returns an error code of 21, the track tables of the R1 and R2 devices need to be merged before you can resume the RDF links. Commands that merge and resume are:

1. symrdf g Rdf1Grp merge symrdf g Rdf1Grp resume

2. symrdf g Rdf1Grp resume force

3. symrdf g Rdf1Grp establish

For a device group with many SRDF pairs, SYMCLI uses these commands to determine which track tables need to be merged and acts only on those tables.

3 To accept symforce with an SRDF command, SYMCLI must recognize that you have enabled -symforce in the options

file. It is recommended that you do not enable this option except for an emergency.


4/5/2004

Establishing an SRDF Pair Establishing an SRDF pair initiates remote mirroring the copying of data from the source (R1) device to the target (R2) device. SRDF pairs come into existence in two different ways:

At configuration time through the pairing of SRDF devices. This is a static pairing configuration.

Any time during Symmetrix operation through your own pairing of dynamic, non-SRDF devices. This is a dynamic pairing configuration in which SRDF pairs can be created and deleted on the fly. See the section Creating SRDF Pairs from Non-SRDF Devices for a description of this feature.

A full establish (symrdf establish full) is typically done when an SRDF pair is initially configured from SRDF devices and connected via the RDF links. Otherwise, you can perform an incremental establish, where the R1 device copies to the R2 device only the new data that was updated while the SRDF pair was split or suspended. To initiate an establish operation on all SRDF pairs in a device group or composite group, all pairs must be in the Split or Suspended state. Use the symrdf query command to check the state of SRDF pairs in a group.

Figure 2 illustrates establishing an SRDF pair. When you initiate the establish operation, the system write-disables the R2 device to its host and performs other tasks. When the establish operation is complete, the SRDF pair is in the Synchronized state. The R1 device and R2 device contain identical data.

SymmetrixSite A

SymmetrixSite B

CLI-000105

Host Host

Source(R1)

Device

Target(R2)

DeviceRDF Links

Write Disabled

R1 data copied to R2

Figure 2. Establishing an SRDF Pair

You can initiate the establish operation from either host, provided that a host has the appropriate device group or composite group definition. The following command initiates an incremental establish operation for all SRDF pairs in the device group named Rdf1Grp:

symrdf g Rdf1Grp establish


4/5/2004

Splitting an SRDF Pair When you need to have read/write access to a target (R2) device, you can split the SRDF pair. When the split completes, the target host can access the R2 device. The R2 device contains valid data and is available for business continuance tasks or restoring data to the R1 device if there is a loss of data on that device.

While an SRDF pair is in the Split state, local I/O to the R1 device can still occur. While these updates are not sent to the R2 device immediately, the system does propagate these updates to the R2 side once you initiate a resumption of normal SRDF mirroring activity.

To initiate a split, an SRDF pair must already be in one of the following states:

Synchronized

Suspended

R1 Updated

SyncInProg and the symforce option was specified4

You can initiate the split operation from either host. The following command initiates a split operation on all SRDF pairs in the device group named Rdf1Grp:

symrdf g Rdf1Grp split

The symrdf split command provides exactly the same functionality as the symrdf suspend and rw enabled R2 commands together. Moreover, the split operation and the suspend operation have exactly the same consistency characteristics for device groups (which use a single RA group for a single Symmetrix unit) and RDF composite groups (which can have multiple RA groups that can span multiple Symmetrix units).

When SRDF pairs are in a device group on a single Symmetrix unit, you can split the SRDF pairs as shown above and have copies on the R2 devices. If the R2 devices need to be consistent (that is, a restartable copy on the R2 side), include the SRDF pairs in a composite group and enable the group for consistency protection.

When a set of SRDF pairs spans multiple Symmetrix units, you can include the SRDF pairs in a composite group and split the group. If consistency is required, enable the composite group for consistency protection. For information on RDF composite groups and consistency, refer to the white paper Using SYMCLI to Implement RDF Consistency Protection with SRDF Family Products (P/N 300-000-284).

Note: It is possible after a split operation that one or more R1 devices may not be mapped to an SA. If so, and if you do not intend to restore R2 data changes to the R1 devices, you should perform an establish operation on these unmapped R1 devices and not a failback operation.

4 To accept symforce with an SRDF command, SYMCLI must recognize that you have enabled -symforce in the options

file. It is recommended that you do not enable this option except for an emergency.


4/5/2004

Restoring an SRDF Pair When you need to copy data from the target (R2) device back to the source (R1) device, you can restore the SRDF pair. After an SRDF pair is split, the R2 device contains valid data and is available for business continuance tasks (such as running a new application) or restoring data to the R1 device if there is a loss of data on that device. Moreover, if the results of running a new application on the R2 device are good, you may want to move the changed data and new application to the R1 device.

You can perform a full restore or incremental restore. A full restore operation copies the entire contents of the R2 device to the R1 device. An incremental restore operation is much quicker because it copies only new data that was updated on the R2 device while the SRDF pair was split. If any tracks on the R1 device changed while the SRDF pair was split, those tracks are refreshed from the corresponding tracks on the R2 device.

To initiate a restore, an SRDF pair must already be in the Split state.

You can initiate the restore operation from either host. The following command initiates an incremental restore operation on all SRDF pairs in the device group named Rdf1Grp (add the full option if you need a full restore):

symrdf g Rdf1Grp restore

The restore operation is complete when the R1 and R2 devices contain identical data. The SRDF pair is then in a Synchronized state.

Failover and Failback Having a synchronized SRDF pair allows you to switch data processing operations from the source site to the target site if operations at the source site are disrupted or if you need to schedule downtime for maintenance. This switchover from source to target is called failover. When operations at the source site are back to normal, you can use a failback operation to re-establish I/O communications links between source and target, resynchronize the data between the sites, and resume normal operation.

In a period of scheduled downtime for maintenance, or after a serious system problem that renders either the source (R1) host or Symmetrix unreachable, no read/write operations can occur on the R1 device. In this case, you can initiate a failover operation to make the paired R2 device read/write enabled to its host.

You can initiate the failover operation from either host. However, if the R1 host is down, the only alternative is to initiate failover from the R2 host. The following command initiates a failover on all SRDF pairs in the Rdf1Grp device group:

symrdf g Rdf1Grp failover

To initiate a failover, the SRDF pair must already be in one of the following states:

Synchronized

Suspended

R1 Updated

Partitioned (when you are invoking this operation from the target side)

Figure 3 illustrates the failover operation. If the R1 device is operational, the RDF links are suspended. If the source side is operational, the R1 device is write disabled to its host. The R2 device is then read/write enabled to its host.


4/5/2004

RDF Links

CLI-000106

Host Host

SymmetrixSite A

SymmetrixSite B

Source(R1)

Device

Target(R2)

Device

Write Disabled

While R1 is unreachable,R2 is read/write enabled to its host.

Figure 3. Failover from the Source System to the Target System

Failback

When you are ready to resume normal SRDF operations, you can initiate a failback (R1 device takeover). This means starting read/write operations on the R1 device, and stopping read/write operations on the R2 device. The R2 becomes read-only to its host, while the R1 becomes read/write enabled to its host.

symrdf g Rdf1Grp failback

To initiate a failback, the SRDF pair must already be in one of the following states:

Failed Over

Suspended and Write Disabled at the source

Suspended and Not Ready at the source

R1 Updated

R1 UpdInProg

Failover with a Fast Swap and Establish

Beginning with Solutions Enabler version 5.4 and Enginuity version 5567, SRDF provides a failover option for dynamic RDF devices that allows you to failover, swap R1/R2 personalities, and establish the dynamic R1 and R2 devices. This operation is useful if you want to transfer data processing to the remote site but continue to replicate this data at the local site. For example:

symrdf g Rdf1Grp failover -establish


4/5/2004

If the device pairs have the same data, the fast swap feature performs the swap and makes the R1/R2 devices read/write (RW) on the link without having to perform the time-consuming merge operation.

If you want to return to the original setup, perform the failover establish operation again from the local host (or from the remote host if device group Rdf1Grp is also defined on the remote host).

Updating the R1 Device You usually perform the R1 update action after a failover, when data processing on the target side has created large numbers of changed tracks on the target devices since the point of the failover. An update operation can bring the R1 device closer to synchronization with the R2 before a failback is initiated. This close-to-synchronized state helps to minimize downtime during the failback process.

You can use the symrdf update command with or without the until option. If you omit the -until option, the system takes a snapshot of whatever tracks have changed on the R2 side (R1 invalid tracks). These tracks are marked to be copied over to the R1 side. If there are new changes to the tracks on the R2 side while the marked tracks are being copied, those changes accumulate as R1 invalid tracks on the R2 side but are not marked for copying. Once the original set of invalid tracks has been copied, the update operation stops.

If you use the until track_threshold option, the system examines the number of R1 invalid tracks on the R2 side once the initial set of tracks has been copied. If the R1 invalid tracks on the R2 side are under the threshold, the update command exits with a successful completion. If the R1 invalid tracks on the R2 side are equal to or greater than the threshold, the update process begins again with a fresh snapshot. This process repeats until the R1 invalid tracks on the R2 side are under the threshold when the R1 update completes.

To initiate an R1 update, the SRDF pair must already be in one of the following states:

R1 Updated

Failed Over

Suspended and Write Disabled at the source

Suspended and Not Ready at the source

Caution: If you perform an update while the SRDF pair is in the Suspended/Not Ready state, the SRDF pair enters an Invalid state as the update completes. To put the SRDF pair in a Synchronized state, you can perform the symrdf rw_enable r1 control operation to write-enable the R1 device to its host.

The following command initiates an update on all R1 devices in the Rdf1Grp device group:

symrdf g Rdf1Grp update

Using symrdf update until # identifies the number of R1 invalid tracks that can accumulate on the R2 side before another update (R2-to-R1 copy) is re-triggered. For example, to update the R1 device until the number of R1 invalid tracks on the R2 side is less than 1000, enter the following command:

symrdf g Rdf1Grp update until 1000

An update sequence begins with an immediate update once the command is initiated. With this operation, a new R1 update cycle will occur every time the previous batch of invalid tracks that was updated has been fully copied to the R1 side and the count of R1 invalid tracks that have accumulated in the interim is equal to or greater than 1000. When the R1 invalid track count at the end of an update cycle is under 1000, no more update cycles are performed. The R1 update is complete.


4/5/2004

Using Concurrent RDF Devices You can have two identical remote copies available at any point in time by having two target R2 devices configured as concurrent remote mirrors of one source R1 device. Using a concurrent SRDF pair like this allows you to easily generate two copies of the same data at remote locations. When you split the two R2 devices from their source R1 device, each target sites host can access its own data at the time of the split.

Concurrent RDF requires two different RDF (RA) groups to achieve the connection between the local R1 device and its two remote R2 mirrors. It is recommended that the RDF groups be on two different RA adapter interfaces, but this is not necessary if you are using Fibre RA adapters. As shown in Figure 4, each RDF (RA) group represents an established link between two Symmetrix units.

CLI-000107

SymmetrixRemote Site A

BCV

R2

SymmetrixRemote Site B

R2RA Group 2RA Group 1

SymmetrixLocal Site

Host

R1

Figure 4. Using RDF Links to Copy the Same Data Concurrently to Two Different Remote Sites

Each of the two remote mirrors can operate with the same method of replication (Synchronous, Semi-Synchronous, Asynchronous, or Adaptive Copy) or in different methods with one exception: concurrent RDF cannot have one remote mirror with Synchronous replication and the other with Semi-Synchronous replication.

When using concurrent RDF, you can build a device group containing concurrent RDF standard devices that belong only to the two RDF groups representing the concurrent links. Your device group can also include BCVs and RDF standard devices that are not concurrent RDF devices. However, within the context of the device group, you can remotely associate a BCV with only one of the concurrent R2 mirrors (as the illustration shows), but not with both5.

5 If you disassociate the BCV at Site A from the device group, you can then remotely associate a BCV from Site B and create a BCV

pair with the concurrent R2 mirror there. However, the BCV pair at Site A is no longer under the control of the device group, even though that BCV pair remains synchronized if the pair was in this state when disassociated from the device group.


4/5/2004

To determine which devices on a particular Symmetrix system have been configured as concurrent RDF devices, use symrdf list with the concurrent option. For example, on Symmetrix 000000003265:

symrdf list sid 3265 concurrent

You can establish a concurrent SRDF pair simultaneously with one command. For example, to incrementally establish the concurrent SRDF pair:

symrdf -g Rdf1Grp establish RDFG ALL

The RDFG ALL option tells SYMCLI to establish the R1-to-R2 links for both RDF groups at the same time. If you want to establish one mirror of the SRDF pair first and then the other, use two establish operations:

symrdf -g Rdf1Grp establish RDFG 1 symrdf -g Rdf1Grp establish RDFG 2

The example assumes that one link of the concurrent SRDF pair is represented by RDF group 1, and the other by RDF group 2.

You can use the symrdf verify command to verify the state of one or both mirrors of the concurrent SRDF pair. For example, to verify the Synchronized state of both concurrent mirrors:

symrdf -g Rdf1Grp verify RDFG ALL

You can also split the concurrent SRDF pair either simultaneously (-RDFG ALL) or one at a time (-RDFG #). The following command splits the concurrent mirror that is represented by RDF group 2:

symrdf -g Rdf1Grp split RDFG 2

The following symrdf query command displays the status of both mirrors of the concurrent SRDF pair:

symrdf -g Rdf1Grp query RDFG ALL

If you split only one of the concurrent mirrors, the link for the split mirror will go to the Not Ready status, and the link for the still-synchronized mirror will remain in the Ready status.

If you want to restore data from the target (R2) devices to the source (R1) device, only one of the concurrent R2 mirrors does the restoring at any given time. (This rule applies to failback and R1 update operations also.) For example, assuming that both concurrent R2 mirrors are split from their R1 standard device, the following command restores the R1 standard device from the R2 mirror represented by RDF group 2:

symrdf -g Rdf1Grp restore RDFG 2

After the restore operation, the R2 mirror associated with RDF group 2 is synchronized with the source (R1) device, and the R2 device associated with RDF group 1 is still in the Split state. If you want the split R2 device to mirror the standard device again, you can simply re-establish them explicitly.

However, if you have written new data to the split R2 device in the interim and you want the split mirrors data to become the resychronized data, you can restore again from the split mirror. In this case, however, you need to include the remote option on the command line. For example:

symrdf -g Rdf1Grp restore RDFG 1 -remote

This operation copies data from the split R2 device to the source (R1) device, and from the R1 device to the R2 mirror that was previously used to restore the R1 device.

For all restore, R1 update, and failback operations where one of the concurrent mirrors is synchronized with its R1 device, using the remote option tells SYMCLI that you intend to copy data from the unsynchronized concurrent mirror to both the R1 device and the other (synchronized) concurrent R2 mirror. If you inadvertently omit the remote option in this case, SYMCLI returns an error message.


4/5/2004

Swapping R1 Devices with R2 Devices In a device group with SRDF pairs, you can swap the R1/R2 personality of all standard RDF or BCV RDF pairs in the device group. Source R1 devices become target R2 devices, and vice versa. Swapping SRDF devices allows the R2 side to take over operations while retaining a remote mirror on the R1 side. Swapping is especially useful after failing over an application from the R1 side to the R2 side.

Before you can perform an R1/R2 swap, all SRDF pairs in the device group must be in one of the following states: Failed Over, R1 Updated, or Suspended. Table 3 illustrates pre- and post-swap states.

Table 3. Pre- and Post-Swap States

SRDF Pair State before the Swap

SRDF Pair State after the Swap

State of New R1 after the Swap

State of New R2 after the Swap

Operation Needed to Resume Mirroring6

Failed Over Suspended Read/Write Enabled Write Disabled Incremental Establish; or Resume

R1 Updated Suspended Read/Write Enabled Write Disabled Incremental Establish; or Resume

Suspended with R1 Write Disabled

Suspended Write Disabled Write Disabled Incremental Establish; or Resume and rw_enable R1

After the swap operation is complete, an SRDF pair is in the Suspended state. If the swap followed a failover or R1 update operation, I/O to the new R1 device is not interrupted; the new R1s state is now that of the pre-swap R2 (read/write enabled). If the swap follows a Suspend operation, the state of the new R1 device after the swap is write disabled. I/O to the new R1 cannot begin until you perform an incremental establish operation.

After the swap, the new R2 device is write disabled, and it can begin functioning as the remote mirror. To begin propagating data from the new R1 to the new R2, perform a symrdf establish operation.

The symrdf swap command can swap all the SRDF devices in the device group, both RDF standard devices and RDF BCVs, as long as the RDF BCV devices belong to the same RDF group as the RDF standard devices. For example, after a failover operation:

symrdf g Rdf1Grp swap all

Omitting the all option swaps just the standard devices, and substituting bcv for all swaps just the RDF BCV devices.

The symrdf establish command re-synchronizes the SRDF pairs incrementally and resumes mirroring between them, albeit in the opposite direction from before the swap operation:

symrdf g Rdf1Grp establish

If you use the refresh R1 option with the swap operation, SYMCLI marks any modified tracks on the pre-swap R1 device to refresh from data on the R2 device. The refresh R2 option does the opposite, but using this option is possible only when the SRDF pair is in the Suspended state prior to the swap.

6 The use of symrdf resume after an R1/R2 swap requires that you either include the force option or that you issue the symrdf merge command prior to the resume operation. SYMCLI uses the force option to initiate the merge operation if Enginuity requests it, and then the resume operation is performed.


4/5/2004

The following command swaps the SRDF standard devices in the device group and marks any modified tracks on the pre-swap R1 device to refresh from data on the R2 device:

symrdf g Rdf1Grp swap refresh R1 i 30 c 10

To execute the swap operation successfully, SYMCLI needs to acquire an exclusive lock on the Symmetrix host database, the local Symmetrix, and the remote Symmetrix systems. If some other operation currently holds this exclusive lock, the previous command tells SYMCLI to retry the swap operation every 30 seconds until the operation succeeds or until it has tried 10 times without success.

With Enginuity version 5568 and the introduction of dynamic SRDF pairs, the static SRDF pairs can no longer be configured for dynamic RDF swap. With version 5568 and higher, the only devices capable of dynamic RDF swap are dynamic SRDF pairs.7 You can create dynamic SRDF pairs using the symrdf createpair command (see the section Creating SRDF Pairs from Non-SRDF Devices).

Dynamic RDF swap is not supported for concurrent dynamic RDF.

Data Mobility Replication Data Mobility replication is an SRDF configuration that restricts SRDF devices to operating only in Adaptive Copy replication. There are two types of Adaptive Copy replication:

Adaptive Copy Disk (AD) replication (acp_disk)

Adaptive Copy Write-Pending replication (acp_wp)

Adaptive Copy replication facilitates data sharing and migration. These methods of replication allow the source and target devices to be more than one I/O out of synchronization. Both methods allow write tasks to accumulate on the local side before being copied to the remote side.

With Adaptive Copy Disk replication, write tasks are stored as invalid tracks on the source device of the SRDF pair. With Adaptive Copy Write-Pending replication, write tasks accumulate in a local cache. A background process moves (or destages) the write-pending tasks to the source device and its corresponding target device. The advantage of this method is that it is typically faster to read data from cache than from disk. The disadvantage is that cache is temporarily consumed by the data until it moves to disk.

At configuration time, EMC can configure SRDF devices for Adaptive Copy replication. By also configuring these devices for Data Mobility replication, EMC applies the restriction. If you attempt to take a restricted device out of Adaptive Copy replication, SYMCLI returns a message to the effect that switching out of Adaptive Copy replication is not possible with data mobility enabled.

To determine if your Symmetrix system has data mobility enabled, you can use the symcfg list v command as shown in the Examples section of this paper.

With data mobility enabled, SYMCLI does not require you to use the force option with a split or failover operation as you would normally need to do when data mobility is not enabled.

7 Devices intended for dynamic RDF swap must be configured with the dyn_rdf attribute, which makes a device capable of being

both a dynamic R1 device and a dynamic R2 device (refer to the white paper Using the SYMCLI Configuration Manager, P/N 300-000-475).


4/5/2004

SRDF Asynchronous (SRDF/A) Replication If a Symmetrix unit is configured to operate with SRDF/A replication8, you can create a device group containing R1 devices that copy data to R2 devices using an all-or-nothing paradigm similar to composite groups. Beginning with Solutions Enabler version 5.3 and Enginuity version 5670, you can use an SRDF/A device group (one containing devices capable of operating with SRDF/A replication) to transfer R1 source data in cycled chunks to the target R2 system. When a complete chunk of data is received on the R2 side, a copy cycle is committed. If the RDF links are lost during data transfer, any partial chunk is discarded, preserving consistency on the R2 with data that is two cycles or less behind the R1.

You can set up and enable SRDF/A such that SRDF devices of an SRDF/A device group act in unison to maintain the integrity of a database being remotely mirrored on a target Symmetrix unit. If a source R1 device in the device group cannot propagate data to its corresponding target R2 device, SRDF/A suspends data propagation from all R1 devices in the device group, halting all data flow to the R2 targets. This suspension of all data propagation, called tripping the device group, ensures a consistent R2 copy of the database up to the point in time that the group was tripped. Tripping an SRDF/A device group can occur either automatically or manually.

An automatic trip occurs when one or more R1 source devices in an enabled SRDF/A device group cannot propagate data to their corresponding R2 target devices. For example:

All RDF links between the R1 and R2 might go down for an extended period of time.

The RDF directors on the R1 side or R2 side might fail.

A manual trip occurs when you invoke the symrdf suspend, split, or failover commands for the SRDF/A device group. Tripping the device group creates a consistent point-in-time R2 image from the N-2 cycle (the cycle that is two cycles behind the R1 side). The current minimum cycle time is 30 seconds.

There are two choices with a manual trip: the default or the immediate option. By default, the SRDF/A session is dropped at the end of the current in-process cycle, which may cause execution time of this command to be longer but results in no invalid tracks on the R2 side. By dropping on a cycle boundary, there is no need to resolve invalid tracks when SRDF/A is resumed.

Dropping the SRDF/A session immediately may result in remote invalid tracks on both the R1 and R2 sides. The SRDF/A devices go NR on the link, and write-pending tracks are converted to invalid tracks on both the R1 and R2 sides. However, dropping a session immediately does not compromise the consistency of the data on the R2 side; SRDF/A always provides a consistent image of the data at the remote site at all times. It is only during a resynchronization operation that data consistency is not guaranteed. (It is recommended that you use BCV devices to preserve a copy of the R2 devices prior to resynchronization.)

The following command drops the SRDF/A session for the device group AsyncGrp. The force option is required here to ensure that you want to stop SRDF/A operation and end consistency protection.

symrdf g AsyncGrp suspend -force

8 SRDF/A replication can be enabled on only one RA (RDF) group in the Symmetrix unit, either using the Symmetrix configuration

server or the Configuration Manager (refer to the white paper Using the SYMCLI Configuration Manager, P/N 300-000-475). All R1 or R2 devices belonging to the RA group enabled for SRDF/A replication become devices capable of participating in SRDF/A operations.


4/5/2004

SRDF/A session status can be one of the following:

Active with consistency protection enabled

The R2 side can be either consistent or inconsistent. If there are any invalid tracks to be copied, the R2 side is not consistent until the N-2 cycle containing the last invalid track is fully on the R2 side.

Active with consistency protection disabled

The SRDF/A devices are ready or NR on the RDF link and running with SRDF/A replication.

Inactive

The SRDF/A devices are ready on the RDF link and are working with their basic methods of replication (Synchronous, Semi-Synchronous, or Adaptive Copy).

To ensure consistency protection, Asynchronous replication must be set and consistency protection enabled on the device group containing the SRDF/A devices. For example:

symrdf g AsyncGrp set mode async symrdf g AsyncGrp enable

Disabling the SRDF/A-backed device group causes consistency protection to be disabled. For example:

symrdf g AsyncGrp disable

SRDF/A devices remain ready on the RDF link and operating with their last primary method of replication. Data continues to flow.

For restrictions regarding the use of SRDF/A replication, refer to EMC Solutions Enabler Symmetrix SRDF CLI Product Guide and to the product release notes.

Using BCVs to Preserve a Copy of the Remote SRDF/A Data

Although SRDF/A replication does not require TimeFinder/Mirror software during normal operation, you may find it useful to mirror the R2 data on TimeFinder BCVs at the remote site to preserve a consistent image of the data on the R2 devices before resynchronization operations. SRDF/A allows you to perform a consistent split on R2-side BCVs without dropping the RDF links, allowing you to preserve point-in-time copies of the R2 data without disrupting the SRDF/A operation between the R1 and R2 devices.

You can control the R2-side BCVs through the RDF1 type device group defined on the local host or an RDF2 type device group defined on the remote host. Controlling remotely-associated BCVs from the R1 side requires using the rdf option. Controlling from the R2 side omits the option.

From the R1-side host, the symmir split rdf command with the consistent option splits the R2-side BCV pairs, making the BCV data available to the R2-side host. For example:

symmir -g AsyncGrp split rdf -consistent

For more information on consistent split, refer to the white paper Using SYMCLI to Perform Consistent Splits with TimeFinder Family Products (P/N 300-000-283).


4/5/2004

Setting Up SRDF Asynchronous Operation

The following steps outline the procedure for performing SRDF control operations with SRDF Asynchronous (SRDF/A) replication (for outputs to these commands, refer to Example 5):

1. List SRDF/A devices on the source Symmetrix unit. The RDF devices displayed belong to the RDF (RA) group that has been configured for SRDF/A operation. The devices displayed will be all R1 devices.

symrdf list -rdfa

2. Create an RDF1 type device group. For example an group named AsyncGrp1:

symdg create AsyncGrp1 type rdf1

3. Add to the device group all devices from the RDF (RA) group configured for SRDF/A operation. For example, if the RDF group displayed in the symrdf list display is group number 1, then all devices in this RDF group must be managed for SRDF/A operation.

symld g AsyncGrp1 addall rdfg 1

4. Query the device group to display the R1-to-R2 setup and the state of the SRDF/A device pairs:

symrdf g AsyncGrp1 query -rdfa

5. Set the device group to Asynchronous replication:

symrdf g AsyncGrp1 set mode async

6. If the SRDF pairs are not in the Consistent state at this time (for example, the Split or Suspended state with invalid tracks on the R1 side), you can initiate SRDF copying of R1 data to the R2 side. The device state will be SyncInProg until the Consistent state is reached.

symrdf -g AsyncGrp1 establish

7. Enable consistency protection for the SRDF/A devices in the device group:

symrdf g AsyncGrp1 enable

8. If you need to manually trip the device group, you can suspend the RDF links for the device group or split the SRDF pairs. By default, the SRDF/A session is dropped at the next switch in copy cycles. The SRDF/A session becomes inactive. The force option is required to ensure that you actually want to stop SRDF/A operation and end consistency protection.

symrdf g AsyncGrp1 suspend -force

9. If there are writes to the R1 source devices while the SRDF/A session is inactive, invalid R1 tracks will accumulate on the R1 side. These are tracks that are owed to the R2 side. If the devices are made ready on the link, you can resume SRDF operation, copying these tracks to the R2 side. The SRDF/A session is automatically activated.

symrdf g AsyncGrp1 resume

10. At this point, the SRDF/A devices are ready on the RDF link and operating with Asynchronous replication. However, consistency protection is not automatically enabled upon resumption of the link. You must explicitly enable consistency protection again.

symrdf g AsyncGrp1 enable


4/5/2004

Performing SRDF Control Operations in Parallel Some SRDF operations require a considerable amount of time to complete, especially establish operations that need to mark and merge track tables between the R1 and R2 devices. If an application such as batch replication to copy data warehouse loads requires establishing device groups across multiple Symmetrix units, doing so sequentially can take a considerable amount of time.

SYMCLI provides a configuration database on your host for SYMAPI access, which by default is set to EXCLUSIVE access to prevent the database from being changed while an operation is in progress. Under this access setting, one establish operation must complete before another begins.

However, if you know that there are no Symmetrix configuration changes currently being made to your hosts SYMAPI database, you can set access to the database to PARALLEL. With parallel access set, SYMCLI can process multiple SRDF operations such as symrdf establish simultaneously. For example, if you need to establish SRDF pairs on three Symmetrix units, you can build a device group for each Symmetrix unit, set the SYMCLI_CTL_ACCESS environment variable to PARALLEL, and issue three symrdf establish commands simultaneously:

Export SYMCLI_CTL_ACCESS=PARALLEL symrdf g group1 establish noprompt & symrdf g group2 establish noprompt & symrdf g group3 establish noprompt & wait

Beginning with the SRDF component of Solutions Enabler version 5.2 running on Symmetrix units using Enginuity version 5667, parallel processing is controlled at the device level rather than at the RDF group level. You can perform up to 64 parallel processing operations on different devices but no more than one operation on any one device at a time. Beginning with Enginuity version 5669, you can perform unlimited parallel processing operations on different devices. Moreover, these operations can span RDF groups.

However, if your Symmetrix unit is still running Solutions Enabler version 5.1 (or lower) with Enginuity version 5568 (or lower) and you want to perform SRDF parallel processing for different RDF groups within a single Symmetrix unit, you need to set the following parameter in the SYMAPI options file:

SYMAPI_PARALLEL_RA_GROUPS = ENABLE

The options file in the SYMAPI configuration directory contains behavior parameters that you can set to change the default behavior of SYMCLI operations. The default for PARALLEL_RA_GROUPS is DISABLE. In the example above, if the three device groups contain devices from three different RDF groups, then these devices can belong to the same Symmetrix unit and can be established using the same parallel processing syntax.


4/5/2004

Creating Dynamic SRDF Pairs When a Symmetrix unit is configured, some devices are usually configured as SRDF devices (an R1 paired with an R2), and others as non-SRDF devices. Dynamic RDF technology allows you to create additional SRDF pairs from non-SRDF devices that have been configured as RDF-capable devices, provided that the Symmetrix unit itself has been enabled for dynamic RDF mode of operation9. You synchronize and control dynamic SRDF pairs the same way that you manage configured SRDF pairs.

Prior to Enginuity version 5568 (as described in Appendix A), source and target SRDF device pairing was limited to those static SRDF pairs set at configuration time. Dynamic RDF provides you with the flexibility to create and delete new SRDF pairs while the Symmetrix unit is in operation.

You can use the symdev list dynamic command to display non-SRDF devices that have been configured as dynamic volumes (see Example 3: Creating Dynamic SRDF Devices). Non-SRDF devices can be configured with the capability to be R1 devices, R2 devices, or both. Once you determine which dynamic devices on the source Symmetrix unit you want to pair with which dynamic devices on the target Symmetrix unit, you need to create a device file and list your device pairs in the file. For example, a list of device pairs in a device file called pairsfile:

09C 054 09D 055 09E 056

Each SRDF pair must be on a separate line in the file (for example, device 09C paired with device 054).

When issuing the symrdf createpair command for this file, specify the device type of first-column devices (R1 or R2 type) and the Symmetrix unit on which the first-column devices reside. Specify also the name of the device file and the RDF (RA) group on the Symmetrix unit that you wish to use as the RDF link between the R1 and R2 devices.

If your initial operation is to establish the pairs, include the establish option or invalidate R2 option. The establish option invalidates the R2 devices, merges the track tables for the pair from both devices, brings up the RDF links, and initiates data copying from the R1 to the R2 devices. If your initial operation is a restore, include the restore option or invalidate R1 option. Including the restore option invalidates the R1 devices, merges the track tables for the pair from both devices, brings up the RDF links, and initiates data copying from the R2 to the R1 devices. The invalidate option allows the creation of dynamic SRDF pairs without bringing up the RDF links and initiating the copying of data.

symrdf createpair file pairsfile sid 77 rdfg 2 type rdf1 invalidate r2

The example executes a file called pairsfile and uses the type rdf1 option to identify the first-column devices as R1 type devices residing on source Symmetrix 000185400077 (sid 77). Therefore, the second-column devices are R2 devices that reside on the target Symmetrix unit. The SRDF pairs can communicate using RDF group number 2 (rdfg 2) from source Symmetrix 000185400077. This pairing information is added to your hosts SYMAPI database file. The invalidate R2 option invalidates the R2 devices in preparation for a subsequent establish operation.

Performing SRDF operations on members of the device file allows you to synchronize new SRDF pairs in the file and query or verify (or both) the progress of the establish or restore operation. For example:

symrdf f pairsfile establish sid 77 rdfg 2 symrdf f pairsfile query sid 77

9 The symcfg list v command displays Symmetrix characteristics, including whether Dynamic RDF Configuration State is

set to enabled.


4/5/2004

To discontinue using these dynamic SRDF pairs, your pairs need to be in a state in which the RDF link state is Not Ready (NR): Suspended, Split, or Failed Over. For example, to suspend the RDF links using a symrdf suspend command, and then perform a symrdf deletepair command:

symrdf f pairsfile suspend sid 77 rdfg 2 symrdf deletepair file pairsfile sid 77 rdfg 2

The symrdf deletepair command cancels the dynamic SRDF pairings by removing the pairing information from your hosts SYMAPI database file.

Beginning with Solutions Enabler version 5.4, if concurrent dynamic RDF is enabled on the Symmetrix unit, you can create concurrent dynamic SRDF pairs by creating a second pairs file that lists the same R1 devices with a different set of R2 devices. For example, a second pairs file called conpairs:

09C 030 09D 031 09E 032

This new set of R2s can be either on the same remote Symmetrix unit as the initial set of R2 devices or on a second remote Symmetrix unit, but you must choose a different RDF group than the one used to connect the initial set of R2 devices. Once you have defined the second pairs file (conpairs), create the concurrent pairs:

symrdf createpair file conpairs sid 77 rdfg 3 type rdf1 invalidate r2

The concurrent SRDF pairs defined in the conpairs file can communicate using RDF group number 3 from the source Symmetrix (sid 77).

Once you have created dynamic SRDF pairs, you can display all, or subsets of, these pairs by omitting or including various options (both, r1, or r2) with the symrdf list dynamic command. Including the both option displays dynamic SRDF pairs in which the paired devices can be either R1 or R2 devices (a requirement for dynamic R1/R2 swap). For example:

symrdf list dynamic -both

Including the r1 option displays only dynamic SRDF pairs in which R1 devices cannot become R2 devices. Including the r2 option displays only pairs in which R2 devices cannot become R1 devices. Omitting all three options displays all dynamic SRDF pairs, regardless of their device configurations.

SRDF allows you to perform SRDF control operations on dynamic SRDF pairs within the more-often-used context of a device group instead of a device file by specifying g on a createpair command line. Thereafter all other SRDF control commands involving the dynamic SRDF pairs can be executed within the context of a device group (for example, DynaGrp). Once the SRDF pairs RDF link state is Not Ready (NR), for instance, you can cancel those dynamic SRDF pairings as follows:

symrdf deletepair -g DynaGrp

This operation changes the type of the device group from RDF1 to Regular. Devices in the device group are changed from R1 devices to standard devices. A symrdf query on the device group returns a message stating that the device group is not an RDF group. A symld list command on the device group (see Example 3) shows that the device group type has changed to Regular and that the same devices that were created as dynamic R1 devices have returned to being RDF-capable standard devices10.

10If you added other devices to the device group (such as pre-configured R1 devices), using symrdf deletepair to cancel the dynamic SRDF pairings in the group may result in an invalid device group. That is, you cannot have R1 devices in a Regular type device group.


4/5/2004

Creating Dynamic RDF Groups An RDF group (RA group) represents a logical connection between two Symmetrix units. Historically, RDF groups were limited to those static RDF groups defined at configuration time. Beginning with Solutions Enabler version 5.2 and Enginuity version 5669, you now have the flexibility to create, modify, and delete RDF groups while the Symmetrix unit is in operation. Currently, the only environment that supports dynamic RDF groups is Switched Fabric SRDF.

An RDF group is a collection of paths (links) between two Symmetrix units. When you create an SRDF pair, that pair is associated with an RDF group to define the communication paths that will be used to synchronize data between the R1 device and R2 device.

A static RDF group is defined using the configuration server and loaded when the Symmetrix unit is configured. This static definition cannot change without using the configuration server to perform the change and load a new configuration.

Creating a dynamic RDF group requires setting the Symmetrix dynamic_rdf parameter in the configuration. The relationship between the pair link and the RDF group is not defined in the configuration but is performed dynamically using host software, allowing you to change these relationships on the fly.

SRDF commands allow you to add, modify (such as adding or deleting RA directors), or delete a dynamic RDF group. Adding a dynamic RDF group creates an empty group. Once the group is created, you can then add dynamic SRDF pairs to it.

The following example creates dynamic RDF group 2 on the local Symmetrix unit and RDF group 3 on the remote Symmetrix unit. (Beginning with Solutions Enabler version 5.2 running on Symmetrix units using Enginuity version 5669, the maximum number of RDF groups is 64.) The command requires that you specify a group label that can be used when modifying or deleting the group.

symrdf addgrp label dynagrp rdfg 2 sid 3264 dir 3A,3B \ -remote_rdfg 3 remote_sid 3265 remote_dir 14A,14B

Creation of RDF group 2 includes directors 3A and 3B from local Symmetrix 3264, and RDF group 3 includes directors 14A and 14B from remote Symmetrix 3265. Before specifying directors, make sure that the physical connections between the local RA and remote RA directors are valid and operational.

The following command adds dynamic SRDF pairs (as defined in the file named pairsfile) to the new dynamic RDF group 2 (as specified by the -rdfg option).

symrdf createpair file pairsfile sid 3264 rdfg 2 type rdf1 invalidate r2

For more information on creating and establishing dynamic SRDF pairs, refer to the previous section.

You can issue the symcfg list ra all switched command to display all RDF groups on the local Symmetrix and its remotely-connected Symmetrix units. The display indicates whether an RDF group is static or dynamic (refer to Example 4: Creating a Dynamic RDF Group).

A dynamic RDF group must be empty to delete it. The following commands delete the SRDF pairs from the group and remove the local and remote dynamic RDF groups that were created using the label dynagrp.

symrdf deletepair file pairsfile sid 3264 rdfg 2 symrdf removegrp sid 3264 label dynagrp

For information about command options and modifying a dynamic RDF group, refer to the EMC Solutions Enabler Symmetrix SRDF CLI Product Guide.


4/5/2004

Using SRDF and TimeFinder/Mirror in Multi-Hop Configurations Using SRDF and TimeFinder/Mirror (also referred to as TimeFinder in this document) to mirror data from a source Symmetrix system to a remotely-associated BCV located across the hall or across the globe creates a global BCV concept. That is, production data can be available on remotely-associated BCVs for data mining, e-business content delivery, application development, testing, backups, and numerous other uses. Moreover, an SRDF multi-hop configuration (multi-hop mirroring to a third site) provides a recovery solution for component or site failures between remotely-mirrored devices. SRDF reduces backup and recovery costs and significantly reduces recovery time after a disaster.

SRDF data propagation allows you to mirror to a Symmetrix system in a third location only the data that has changed since the last update. By copying only the changed tracks, you consume less bandwidth and enhance performance. You can perform Symmetrix-to-Symmetrix transmissions synchronously in the local or campus area, or use delayed synchronization methods of replication for long-distance transmission. Multi-hop mirroring to a third site can take place during off-peak times or over lower cost transmission lines or via IP-based Virtual Private Networks.

By using SRDF/AR, you can set up remote mirroring on an automatic basis according to your own predefined copy cycle (every hour, for example). For more information about SRDF/AR, refer to the white paper Using SYMCLI to Perform SRDF/AR (P/N 300-000-078).

Establishing a BCV Pair with an RDF BCV

With both SRDF and TimeFinder/Mirror installed, EMC can configure an SRDF device either as an RDF standard device or an RDF BCV device. An RDF BCV device is a BCV that is configured in a one-to-one relationship with a remote target (R2) device that mirrors the BCV, forming an SRDF pair as shown in the illustration. You can also establish an RDF BCV device as part of a BCV pair (shown in Figure 5 on the Site A Symmetrix system) but, in doing so, the RDF BCV is suspended from copying data to its target (R2) device.

SymmetrixSite A

SymmetrixSite B

SRDFPair

Standard

Source(R1)

RDF BCV

BCV Pair

Target(R2)

Standard

CLI-000108

Host

Figure 5. Distinguishing between an SRDF Pair and an RDF BCV Pair


4/5/2004

To copy data from the standard device to the RDF BCV device at Site A, you can build a Regular type device group (RegGrp, for example) on the local host that includes these two devices. For example:

symdg create RegGrp type regular symld g RegGrp add dev 020 symbcv g RegGrp associate dev 090

Use the symmir establish command to initiate their synchronization:

symmir g RegGrp establish -full

When established as a BCV pair, the RDF BCV is suspended from copying data to its remote (R2) target. To resume mirroring between them, you can perform a normal split and re-establish the RDF BCV with its remote (R2) target device as shown in the following example:

symmir g RegGrp split symrdf g RegGrp establish bcv

The symrdf establish command with the bcv option resumes the RDF links and initiates the propagation of data from the source RDF BCV device to its remote (R2) target device.

If later you want to re-establish the BCV pair incrementally, perform a symmir establish. If, instead, you want to resynchronize the BCV pair but restore t

symcli control operations with srdf family products

Documents

srdf configurations

srdf functionality

srdf control operations

srdf pair states

srdf family products

publication date

emc software

emc corporation